In this paper, optimal designs for precast beams were determined using a genetic algorithm. Two new features for enhancing genetic algorithms were developed in the present study, considerable improvements obtained by applying the proposed methods were shown in a parametric study. Firstly, probabilistic-based natural selection was introduced, selecting parental chromosomes for reproduction using inherited probabilities calculated by the ranking of each design among populations. Secondly, mutations using dynamic ranges were proposed, in which, the mutated parameters were varied between a dynamic range around the inherited values instead of being selected randomly. A dynamic range was expanded according to the convergencies of the objective indexes, allowing the application of high mutation rates without degrading searching efficiencies. Furthermore, genetic algorithm-based design charts were constructed, offering reasonable references for preliminary designs. Overall, the proposed procedures showed adequate applications in practical designs, improving the design efficiencies, and reducing human efforts.

在本文中，使用遗传算法确定了预制梁的最佳设计。本研究开发了两个用于增强遗传算法的新功能，参数研究显示了通过应用所提出的方法获得的显着改进。首先，引入了基于概率的自然选择，使用通过种群中每个设计的排名计算的遗传概率来选择用于繁殖的亲本染色体。其次，提出了使用动态范围的突变，其中，突变参数在继承值周围的动态范围之间变化，而不是随机选择。根据目标指标的收敛性扩大了动态范围，允许在不降低搜索效率的情况下应用高突变率。此外，构建了基于遗传算法的设计图，为初步设计提供了合理的参考。总体而言，所提出的程序在实际设计中显示出充分的应用，提高了设计效率，并减少了人力。